#include <unictype.h>
#include <unistd.h>
#include <unistr.h>
-#include <uniwidth.h>
#include "language/command.h"
#include "language/lexer/macro.h"
#include "libpspp/cast.h"
#include "libpspp/deque.h"
#include "libpspp/i18n.h"
+#include "libpspp/intern.h"
#include "libpspp/ll.h"
#include "libpspp/message.h"
#include "libpspp/misc.h"
location of the token in terms of the lex_source's buffer.
For a token produced through macro expansion, this is the entire macro
- call.
-
- src->tail <= line_pos <= token_pos <= src->head. */
- size_t token_pos; /* Start of token. */
+ call. */
+ size_t token_pos; /* Offset into src->buffer of token start. */
size_t token_len; /* Length of source for token in bytes. */
- size_t line_pos; /* Start of line containing token_pos. */
- int first_line; /* Line number at token_pos. */
/* For a token obtained through macro expansion, this is just this token.
size_t *ref_cnt; /* Number of lex_tokens that refer to macro_rep. */
};
+static struct msg_point lex_token_start_point (const struct lex_source *,
+ const struct lex_token *);
+static struct msg_point lex_token_end_point (const struct lex_source *,
+ const struct lex_token *);
+
+/* Source offset of the last byte in TOKEN. */
+static size_t
+lex_token_end (const struct lex_token *token)
+{
+ return token->token_pos + MAX (token->token_len, 1) - 1;
+}
+
static void
lex_token_destroy (struct lex_token *t)
{
static size_t lex_stage_count (const struct lex_stage *);
static bool lex_stage_is_empty (const struct lex_stage *);
-static struct lex_token *lex_stage_last (struct lex_stage *);
static struct lex_token *lex_stage_first (struct lex_stage *);
static struct lex_token *lex_stage_nth (struct lex_stage *, size_t ofs);
return deque_count (&stage->deque);
}
-/* Returns the last token in STAGE, which must be nonempty. The last token is
- the one accessed with the greatest lookahead. */
-static struct lex_token *
-lex_stage_last (struct lex_stage *stage)
-{
- return stage->tokens[deque_front (&stage->deque, 0)];
-}
-
/* Returns the first token in STAGE, which must be nonempty.
The first token is the one accessed with the least lookahead. */
static struct lex_token *
stage->tokens[deque_push_front (&stage->deque)] = token;
}
+/* Removes and returns the first token from STAGE. */
+static struct lex_token *
+lex_stage_take_first (struct lex_stage *stage)
+{
+ return stage->tokens[deque_pop_back (&stage->deque)];
+}
+
/* Removes the first token from STAGE and uninitializes it. */
static void
lex_stage_pop_first (struct lex_stage *stage)
{
- lex_token_destroy (stage->tokens[deque_pop_back (&stage->deque)]);
+ lex_token_destroy (lex_stage_take_first (stage));
}
/* Removes the first N tokens from SRC, appending them to DST as the last
lex_stage_shift (struct lex_stage *dst, struct lex_stage *src, size_t n)
{
for (size_t i = 0; i < n; i++)
- {
- lex_stage_push_last (dst, lex_stage_first (src));
- deque_pop_back (&src->deque);
- }
+ lex_stage_push_last (dst, lex_stage_take_first (src));
}
/* A source of tokens, corresponding to a syntax file.
struct lex_source
{
struct ll ll; /* In lexer's list of sources. */
+
+ /* Reference count:
+
+ - One for struct lexer.
+
+ - One for each struct msg_location that references this source. */
+ size_t n_refs;
+
struct lex_reader *reader;
struct lexer *lexer;
struct segmenter segmenter;
bool eof; /* True if T_STOP was read from 'reader'. */
/* Buffer of UTF-8 bytes. */
- char *buffer;
+ char *buffer; /* Source file contents. */
+ size_t length; /* Number of bytes filled. */
size_t allocated; /* Number of bytes allocated. */
- size_t tail; /* &buffer[0] offset into UTF-8 source. */
- size_t head; /* &buffer[head - tail] offset into source. */
- /* Positions in source file, tail <= pos <= head for each member here. */
+ /* Offsets into 'buffer'. */
size_t journal_pos; /* First byte not yet output to journal. */
size_t seg_pos; /* First byte not yet scanned as token. */
- size_t line_pos; /* First byte of line containing seg_pos. */
- int n_newlines; /* Number of new-lines up to seg_pos. */
+ /* Offset into 'buffer' of starts of lines. */
+ size_t *lines;
+ size_t n_lines, allocated_lines;
+
bool suppress_next_newline;
/* Tokens.
in 'merge'.
- merge: Tokens that need to pass through scan_merge() to end up in
- 'lookahead'.
+ 'parse'.
+
+ - parse: Tokens available to the client for parsing.
- - lookahead: Tokens available to the client for parsing. */
+ 'pp' and 'merge' store tokens only temporarily until they pass into
+ 'parse'. Tokens then live in 'parse' until the command is fully
+ consumed, at which time they are freed together. */
struct lex_stage pp;
struct lex_stage merge;
- struct lex_stage lookahead;
+ struct lex_token **parse;
+ size_t n_parse, allocated_parse, parse_ofs;
};
static struct lex_source *lex_source_create (struct lexer *,
struct lex_reader *);
-static void lex_source_destroy (struct lex_source *);
/* Lexer. */
struct lexer
};
static struct lex_source *lex_source__ (const struct lexer *);
-static char *lex_source_get_syntax__ (const struct lex_source *,
- int n0, int n1);
+static char *lex_source_syntax__ (const struct lex_source *,
+ int ofs0, int ofs1);
static const struct lex_token *lex_next__ (const struct lexer *, int n);
static void lex_source_push_endcmd__ (struct lex_source *);
+static void lex_source_push_parse (struct lex_source *, struct lex_token *);
+static void lex_source_clear_parse (struct lex_source *);
-static bool lex_source_get_lookahead (struct lex_source *);
-static void lex_source_error_valist (struct lex_source *, int n0, int n1,
+static bool lex_source_get_parse (struct lex_source *);
+static void lex_source_error_valist (struct lex_source *, int ofs0, int ofs1,
const char *format, va_list)
PRINTF_FORMAT (4, 0);
static const struct lex_token *lex_source_next__ (const struct lex_source *,
struct lex_source *source, *next;
ll_for_each_safe (source, next, struct lex_source, ll, &lexer->sources)
- lex_source_destroy (source);
+ {
+ ll_remove (&source->ll);
+ lex_source_unref (source);
+ }
macro_set_destroy (lexer->macros);
free (lexer);
}
if (src == NULL)
return;
- if (!lex_stage_is_empty (&src->lookahead))
- lex_stage_pop_first (&src->lookahead);
+ if (src->parse_ofs < src->n_parse)
+ {
+ if (src->parse[src->parse_ofs]->token.type == T_ENDCMD)
+ lex_source_clear_parse (src);
+ else
+ src->parse_ofs++;
+ }
- while (lex_stage_is_empty (&src->lookahead))
- if (!lex_source_get_lookahead (src))
+ while (src->parse_ofs == src->n_parse)
+ if (!lex_source_get_parse (src))
{
- lex_source_destroy (src);
+ ll_remove (&src->ll);
+ lex_source_unref (src);
src = lex_source__ (lexer);
if (src == NULL)
return;
}
}
+
+/* Advances LEXER by N tokens. */
+void
+lex_get_n (struct lexer *lexer, size_t n)
+{
+ while (n-- > 0)
+ lex_get (lexer);
+}
\f
/* Issuing errors. */
va_list args;
va_start (args, format);
- lex_next_error_valist (lexer, 0, 0, format, args);
+ lex_ofs_error_valist (lexer, lex_ofs (lexer), lex_ofs (lexer), format, args);
va_end (args);
}
void
lex_error_valist (struct lexer *lexer, const char *format, va_list args)
{
- lex_next_error_valist (lexer, 0, 0, format, args);
+ lex_ofs_error_valist (lexer, lex_ofs (lexer), lex_ofs (lexer), format, args);
}
-/* Prints a syntax error message containing the current token and
- given message MESSAGE (if non-null). */
+/* Prints a syntax error message for the span of tokens N0 through N1,
+ inclusive, from the current token in LEXER, adding message MESSAGE (if
+ non-null). */
void
lex_next_error (struct lexer *lexer, int n0, int n1, const char *format, ...)
{
va_list args;
va_start (args, format);
- lex_next_error_valist (lexer, n0, n1, format, args);
+ int ofs = lex_ofs (lexer);
+ lex_ofs_error_valist (lexer, n0 + ofs, n1 + ofs, format, args);
+ va_end (args);
+}
+
+/* Prints a syntax error message for the span of tokens with offsets OFS0
+ through OFS1, inclusive, within the current command in LEXER, adding message
+ MESSAGE (if non-null). */
+void
+lex_ofs_error (struct lexer *lexer, int ofs0, int ofs1, const char *format, ...)
+{
+ va_list args;
+
+ va_start (args, format);
+ lex_ofs_error_valist (lexer, ofs0, ofs1, format, args);
va_end (args);
}
options[5], options[6], options[7]);
break;
+ case 9:
+ lex_error (lexer, _("expecting %s, %s, %s, %s, %s, %s, %s, %s, or %s"),
+ options[0], options[1], options[2], options[3], options[4],
+ options[5], options[6], options[7], options[8]);
+ break;
+
default:
lex_error (lexer, NULL);
}
sbc, spec);
}
-/* Prints a syntax error message containing the current token and
- given message MESSAGE (if non-null). */
+/* Prints a syntax error message for the span of tokens with offsets OFS0
+ through OFS1, inclusive, within the current command in LEXER, adding message
+ MESSAGE (if non-null) with the given ARGS. */
void
-lex_next_error_valist (struct lexer *lexer, int n0, int n1,
- const char *format, va_list args)
+lex_ofs_error_valist (struct lexer *lexer, int ofs0, int ofs1,
+ const char *format, va_list args)
{
struct lex_source *src = lex_source__ (lexer);
if (src != NULL)
- lex_source_error_valist (src, n0, n1, format, args);
+ lex_source_error_valist (src, ofs0, ofs1, format, args);
else
{
struct string s;
ds_put_cstr (&s, ": ");
ds_put_vformat (&s, format, args);
}
- ds_put_byte (&s, '.');
+ if (ds_last (&s) != '.')
+ ds_put_byte (&s, '.');
msg (SE, "%s", ds_cstr (&s));
ds_destroy (&s);
}
return false;
}
+/* If the current token is an number in the closed range [MIN,MAX], does
+ nothing and returns true. Otherwise, reports an error and returns false.
+ If NAME is nonnull, then it is used in the error message. */
+bool
+lex_force_num_range_closed (struct lexer *lexer, const char *name,
+ double min, double max)
+{
+ bool is_number = lex_is_number (lexer);
+ bool too_small = is_number && lex_number (lexer) < min;
+ bool too_big = is_number && lex_number (lexer) > max;
+ if (is_number && !too_small && !too_big)
+ return true;
+
+ if (min > max)
+ {
+ /* Weird, maybe a bug in the caller. Just report that we needed an
+ number. */
+ if (name)
+ lex_error (lexer, _("Number expected for %s."), name);
+ else
+ lex_error (lexer, _("Number expected."));
+ }
+ else if (min == max)
+ {
+ if (name)
+ lex_error (lexer, _("Expected %g for %s."), min, name);
+ else
+ lex_error (lexer, _("Expected %g."), min);
+ }
+ else
+ {
+ bool report_lower_bound = min > -DBL_MAX || too_small;
+ bool report_upper_bound = max < DBL_MAX || too_big;
+
+ if (report_lower_bound && report_upper_bound)
+ {
+ if (name)
+ lex_error (lexer,
+ _("Expected number between %g and %g for %s."),
+ min, max, name);
+ else
+ lex_error (lexer, _("Expected number between %g and %g."),
+ min, max);
+ }
+ else if (report_lower_bound)
+ {
+ if (min == 0)
+ {
+ if (name)
+ lex_error (lexer, _("Expected non-negative number for %s."),
+ name);
+ else
+ lex_error (lexer, _("Expected non-negative number."));
+ }
+ else
+ {
+ if (name)
+ lex_error (lexer, _("Expected number %g or greater for %s."),
+ min, name);
+ else
+ lex_error (lexer, _("Expected number %g or greater."), min);
+ }
+ }
+ else if (report_upper_bound)
+ {
+ if (name)
+ lex_error (lexer,
+ _("Expected number less than or equal to %g for %s."),
+ max, name);
+ else
+ lex_error (lexer, _("Expected number less than or equal to %g."),
+ max);
+ }
+ else
+ {
+ if (name)
+ lex_error (lexer, _("Number expected for %s."), name);
+ else
+ lex_error (lexer, _("Number expected."));
+ }
+ }
+ return false;
+}
+
+/* If the current token is an number in the half-open range [MIN,MAX), does
+ nothing and returns true. Otherwise, reports an error and returns false.
+ If NAME is nonnull, then it is used in the error message. */
+bool
+lex_force_num_range_halfopen (struct lexer *lexer, const char *name,
+ double min, double max)
+{
+ bool is_number = lex_is_number (lexer);
+ bool too_small = is_number && lex_number (lexer) < min;
+ bool too_big = is_number && lex_number (lexer) >= max;
+ if (is_number && !too_small && !too_big)
+ return true;
+
+ if (min >= max)
+ {
+ /* Weird, maybe a bug in the caller. Just report that we needed an
+ number. */
+ if (name)
+ lex_error (lexer, _("Number expected for %s."), name);
+ else
+ lex_error (lexer, _("Number expected."));
+ }
+ else
+ {
+ bool report_lower_bound = min > -DBL_MAX || too_small;
+ bool report_upper_bound = max < DBL_MAX || too_big;
+
+ if (report_lower_bound && report_upper_bound)
+ {
+ if (name)
+ lex_error (lexer, _("Expected number in [%g,%g) for %s."),
+ min, max, name);
+ else
+ lex_error (lexer, _("Expected number in [%g,%g)."),
+ min, max);
+ }
+ else if (report_lower_bound)
+ {
+ if (min == 0)
+ {
+ if (name)
+ lex_error (lexer, _("Expected non-negative number for %s."),
+ name);
+ else
+ lex_error (lexer, _("Expected non-negative number."));
+ }
+ else
+ {
+ if (name)
+ lex_error (lexer, _("Expected number %g or greater for %s."),
+ min, name);
+ else
+ lex_error (lexer, _("Expected number %g or greater."), min);
+ }
+ }
+ else if (report_upper_bound)
+ {
+ if (name)
+ lex_error (lexer,
+ _("Expected number less than %g for %s."), max, name);
+ else
+ lex_error (lexer, _("Expected number less than %g."), max);
+ }
+ else
+ {
+ if (name)
+ lex_error (lexer, _("Number expected for %s."), name);
+ else
+ lex_error (lexer, _("Number expected."));
+ }
+ }
+ return false;
+}
+
+/* If the current token is an number in the open range (MIN,MAX], does
+ nothing and returns true. Otherwise, reports an error and returns false.
+ If NAME is nonnull, then it is used in the error message. */
+bool
+lex_force_num_range_open (struct lexer *lexer, const char *name,
+ double min, double max)
+{
+ bool is_number = lex_is_number (lexer);
+ bool too_small = is_number && lex_number (lexer) <= min;
+ bool too_big = is_number && lex_number (lexer) >= max;
+ if (is_number && !too_small && !too_big)
+ return true;
+
+ if (min >= max)
+ {
+ /* Weird, maybe a bug in the caller. Just report that we needed an
+ number. */
+ if (name)
+ lex_error (lexer, _("Number expected for %s."), name);
+ else
+ lex_error (lexer, _("Number expected."));
+ }
+ else
+ {
+ bool report_lower_bound = min > -DBL_MAX || too_small;
+ bool report_upper_bound = max < DBL_MAX || too_big;
+
+ if (report_lower_bound && report_upper_bound)
+ {
+ if (name)
+ lex_error (lexer, _("Expected number in (%g,%g) for %s."),
+ min, max, name);
+ else
+ lex_error (lexer, _("Expected number in (%g,%g)."), min, max);
+ }
+ else if (report_lower_bound)
+ {
+ if (min == 0)
+ {
+ if (name)
+ lex_error (lexer, _("Expected positive number for %s."), name);
+ else
+ lex_error (lexer, _("Expected positive number."));
+ }
+ else
+ {
+ if (name)
+ lex_error (lexer, _("Expected number greater than %g for %s."),
+ min, name);
+ else
+ lex_error (lexer, _("Expected number greater than %g."), min);
+ }
+ }
+ else if (report_upper_bound)
+ {
+ if (name)
+ lex_error (lexer, _("Expected number less than %g for %s."),
+ max, name);
+ else
+ lex_error (lexer, _("Expected number less than %g."), max);
+ }
+ else
+ {
+ if (name)
+ lex_error (lexer, _("Number expected for %s."), name);
+ else
+ lex_error (lexer, _("Number expected."));
+ }
+ }
+ return false;
+}
+
/* If the current token is an identifier, does nothing and returns true.
Otherwise, reports an error and returns false. */
bool
}
static const struct lex_token *
-lex_source_next__ (const struct lex_source *src_, int n)
+lex_source_ofs__ (const struct lex_source *src_, int ofs)
{
struct lex_source *src = CONST_CAST (struct lex_source *, src_);
- while (lex_stage_count (&src->lookahead) <= n)
+
+ if (ofs < 0)
{
- if (!lex_stage_is_empty (&src->lookahead))
+ static const struct lex_token endcmd_token
+ = { .token = { .type = T_ENDCMD } };
+ return &endcmd_token;
+ }
+
+ while (ofs >= src->n_parse)
+ {
+ if (src->n_parse > 0)
{
- const struct lex_token *t = lex_stage_last (&src->lookahead);
+ const struct lex_token *t = src->parse[src->n_parse - 1];
if (t->token.type == T_STOP || t->token.type == T_ENDCMD)
return t;
}
- lex_source_get_lookahead (src);
+ lex_source_get_parse (src);
}
- return lex_stage_nth (&src->lookahead, n);
+ return src->parse[ofs];
+}
+
+static const struct lex_token *
+lex_source_next__ (const struct lex_source *src, int n)
+{
+ return lex_source_ofs__ (src, n + src->parse_ofs);
}
/* Returns the "struct token" of the token N after the current one in LEXER.
return lex_next (lexer, n)->string;
}
+/* Returns the offset of the current token within the command being parsed in
+ LEXER. This is 0 for the first token in a command, 1 for the second, and so
+ on. The return value is useful later for referring to this token in calls
+ to lex_ofs_*(). */
+int
+lex_ofs (const struct lexer *lexer)
+{
+ struct lex_source *src = lex_source__ (lexer);
+ return src ? src->parse_ofs : 0;
+}
+
+/* Returns the token within LEXER's current command with offset OFS. Use
+ lex_ofs() to find out the offset of the current token. */
+const struct token *
+lex_ofs_token (const struct lexer *lexer_, int ofs)
+{
+ struct lexer *lexer = CONST_CAST (struct lexer *, lexer_);
+ struct lex_source *src = lex_source__ (lexer);
+
+ if (src != NULL)
+ return &lex_source_next__ (src, ofs - src->parse_ofs)->token;
+ else
+ {
+ static const struct token stop_token = { .type = T_STOP };
+ return &stop_token;
+ }
+}
+
+/* Allocates and returns a new struct msg_location that spans tokens with
+ offsets OFS0 through OFS1, inclusive, within the current command in
+ LEXER. See lex_ofs() for an explanation of token offsets.
+
+ The caller owns and must eventually free the returned object. */
+struct msg_location *
+lex_ofs_location (const struct lexer *lexer, int ofs0, int ofs1)
+{
+ int ofs = lex_ofs (lexer);
+ return lex_get_location (lexer, ofs0 - ofs, ofs1 - ofs);
+}
+
+/* Returns a msg_point for the first character in the token with offset OFS,
+ where offset 0 is the first token in the command currently being parsed, 1
+ the second token, and so on. These are absolute offsets, not relative to
+ the token currently being parsed within the command.
+
+ Returns zeros for a T_STOP token.
+ */
+struct msg_point
+lex_ofs_start_point (const struct lexer *lexer, int ofs)
+{
+ const struct lex_source *src = lex_source__ (lexer);
+ return (src
+ ? lex_token_start_point (src, lex_source_ofs__ (src, ofs))
+ : (struct msg_point) { 0, 0 });
+}
+
+/* Returns a msg_point for the last character, inclusive, in the token with
+ offset OFS, where offset 0 is the first token in the command currently being
+ parsed, 1 the second token, and so on. These are absolute offsets, not
+ relative to the token currently being parsed within the command.
+
+ Returns zeros for a T_STOP token.
+
+ Most of the time, a single token is wholly within a single line of syntax,
+ so that the start and end point for a given offset have the same line
+ number. There are two exceptions: a T_STRING token can be made up of
+ multiple segments on adjacent lines connected with "+" punctuators, and a
+ T_NEG_NUM token can consist of a "-" on one line followed by the number on
+ the next.
+ */
+struct msg_point
+lex_ofs_end_point (const struct lexer *lexer, int ofs)
+{
+ const struct lex_source *src = lex_source__ (lexer);
+ return (src
+ ? lex_token_end_point (src, lex_source_ofs__ (src, ofs))
+ : (struct msg_point) { 0, 0 });
+}
+
/* Returns the text of the syntax in tokens N0 ahead of the current one,
through N1 ahead of the current one, inclusive. (For example, if N0 and N1
- are both zero, this requests the syntax for the current token.) The caller
- must eventually free the returned string (with free()). The syntax is
- encoded in UTF-8 and in the original form supplied to the lexer so that, for
- example, it may include comments, spaces, and new-lines if it spans multiple
- tokens. Macro expansion, however, has already been performed. */
+ are both zero, this requests the syntax for the current token.)
+
+ The caller must eventually free the returned string (with free()). The
+ syntax is encoded in UTF-8 and in the original form supplied to the lexer so
+ that, for example, it may include comments, spaces, and new-lines if it
+ spans multiple tokens. Macro expansion, however, has already been
+ performed. */
char *
lex_next_representation (const struct lexer *lexer, int n0, int n1)
{
- return lex_source_get_syntax__ (lex_source__ (lexer), n0, n1);
+ const struct lex_source *src = lex_source__ (lexer);
+ return (src
+ ? lex_source_syntax__ (src, n0 + src->parse_ofs, n1 + src->parse_ofs)
+ : xstrdup (""));
+}
+
+
+/* Returns the text of the syntax in tokens with offsets OFS0 to OFS1,
+ inclusive. (For example, if OFS0 and OFS1 are both zero, this requests the
+ syntax for the first token in the current command.)
+
+ The caller must eventually free the returned string (with free()). The
+ syntax is encoded in UTF-8 and in the original form supplied to the lexer so
+ that, for example, it may include comments, spaces, and new-lines if it
+ spans multiple tokens. Macro expansion, however, has already been
+ performed. */
+char *
+lex_ofs_representation (const struct lexer *lexer, int ofs0, int ofs1)
+{
+ const struct lex_source *src = lex_source__ (lexer);
+ return src ? lex_source_syntax__ (src, ofs0, ofs1) : xstrdup ("");
}
/* Returns true if the token N ahead of the current one was produced by macro
}
}
-/* If LEXER is positioned at the sequence of tokens that may be parsed from S,
- skips it and returns true. Otherwise, returns false.
-
- S may consist of an arbitrary sequence of tokens, e.g. "KRUSKAL-WALLIS",
- "2SLS", or "END INPUT PROGRAM". Identifiers may be abbreviated to their
- first three letters. */
-bool
-lex_match_phrase (struct lexer *lexer, const char *s)
+static size_t
+lex_at_phrase__ (struct lexer *lexer, const char *s)
{
struct string_lexer slex;
struct token token;
- int i;
- i = 0;
+ size_t i = 0;
string_lexer_init (&slex, s, strlen (s), SEG_MODE_INTERACTIVE, true);
while (string_lexer_next (&slex, &token))
{
bool match = lex_tokens_match (lex_next (lexer, i++), &token);
token_uninit (&token);
if (!match)
- return false;
+ return 0;
}
-
- while (i-- > 0)
- lex_get (lexer);
- return true;
+ return i;
}
-static int
-count_newlines (char *s, size_t length)
+/* If LEXER is positioned at the sequence of tokens that may be parsed from S,
+ returns true. Otherwise, returns false.
+
+ S may consist of an arbitrary sequence of tokens, e.g. "KRUSKAL-WALLIS",
+ "2SLS", or "END INPUT PROGRAM". Identifiers may be abbreviated to their
+ first three letters. */
+bool
+lex_at_phrase (struct lexer *lexer, const char *s)
{
- int n_newlines = 0;
- char *newline;
+ return lex_at_phrase__ (lexer, s) > 0;
+}
- while ((newline = memchr (s, '\n', length)) != NULL)
- {
- n_newlines++;
- length -= (newline + 1) - s;
- s = newline + 1;
- }
+/* If LEXER is positioned at the sequence of tokens that may be parsed from S,
+ skips it and returns true. Otherwise, returns false.
- return n_newlines;
+ S may consist of an arbitrary sequence of tokens, e.g. "KRUSKAL-WALLIS",
+ "2SLS", or "END INPUT PROGRAM". Identifiers may be abbreviated to their
+ first three letters. */
+bool
+lex_match_phrase (struct lexer *lexer, const char *s)
+{
+ size_t n = lex_at_phrase__ (lexer, s);
+ if (n > 0)
+ lex_get_n (lexer, n);
+ return n > 0;
}
+/* Returns the 1-based line number of the source text at the byte OFFSET in
+ SRC. */
static int
-lex_token_get_last_line_number (const struct lex_source *src,
- const struct lex_token *token)
+lex_source_ofs_to_line_number (const struct lex_source *src, size_t offset)
{
- if (token->first_line == 0)
- return 0;
- else
+ size_t lo = 0;
+ size_t hi = src->n_lines;
+ for (;;)
{
- char *token_str = &src->buffer[token->token_pos - src->tail];
- return token->first_line + count_newlines (token_str, token->token_len) + 1;
+ size_t mid = (lo + hi) / 2;
+ if (mid + 1 >= src->n_lines)
+ return src->n_lines;
+ else if (offset >= src->lines[mid + 1])
+ lo = mid;
+ else if (offset < src->lines[mid])
+ hi = mid;
+ else
+ return mid + 1;
}
}
+/* Returns the 1-based column number of the source text at the byte OFFSET in
+ SRC. */
static int
-count_columns (const char *s_, size_t length)
+lex_source_ofs_to_column_number (const struct lex_source *src, size_t offset)
{
- const uint8_t *s = CHAR_CAST (const uint8_t *, s_);
- int columns;
- size_t ofs;
- int mblen;
-
- columns = 0;
- for (ofs = 0; ofs < length; ofs += mblen)
- {
- ucs4_t uc;
-
- mblen = u8_mbtouc (&uc, s + ofs, length - ofs);
- if (uc != '\t')
- {
- int width = uc_width (uc, "UTF-8");
- if (width > 0)
- columns += width;
- }
- else
- columns = ROUND_UP (columns + 1, 8);
- }
-
- return columns + 1;
+ const char *newline = memrchr (src->buffer, '\n', offset);
+ size_t line_ofs = newline ? newline - src->buffer + 1 : 0;
+ return utf8_count_columns (&src->buffer[line_ofs], offset - line_ofs) + 1;
}
-static int
-lex_token_get_first_column (const struct lex_source *src,
- const struct lex_token *token)
+static struct msg_point
+lex_source_ofs_to_point__ (const struct lex_source *src, size_t offset)
{
- return count_columns (&src->buffer[token->line_pos - src->tail],
- token->token_pos - token->line_pos);
+ return (struct msg_point) {
+ .line = lex_source_ofs_to_line_number (src, offset),
+ .column = lex_source_ofs_to_column_number (src, offset),
+ };
}
-static int
-lex_token_get_last_column (const struct lex_source *src,
- const struct lex_token *token)
+static struct msg_point
+lex_token_start_point (const struct lex_source *src,
+ const struct lex_token *token)
{
- char *start, *end, *newline;
+ return lex_source_ofs_to_point__ (src, token->token_pos);
+}
- start = &src->buffer[token->line_pos - src->tail];
- end = &src->buffer[(token->token_pos + token->token_len) - src->tail];
- newline = memrchr (start, '\n', end - start);
- if (newline != NULL)
- start = newline + 1;
- return count_columns (start, end - start);
+static struct msg_point
+lex_token_end_point (const struct lex_source *src,
+ const struct lex_token *token)
+{
+ return lex_source_ofs_to_point__ (src, lex_token_end (token));
}
static struct msg_location
const struct lex_token *t1)
{
return (struct msg_location) {
- .file_name = src->reader->file_name,
- .first_line = t0->first_line,
- .last_line = lex_token_get_last_line_number (src, t1),
- .first_column = lex_token_get_first_column (src, t0),
- .last_column = lex_token_get_last_column (src, t1),
+ .file_name = intern_new_if_nonnull (src->reader->file_name),
+ .start = lex_token_start_point (src, t0),
+ .end = lex_token_end_point (src, t1),
};
}
}
static struct msg_location *
-lex_source_get_location (const struct lex_source *src, int n0, int n1)
+lex_source_get_location (const struct lex_source *src, int ofs0, int ofs1)
{
return lex_token_location_rw (src,
- lex_source_next__ (src, n0),
- lex_source_next__ (src, n1));
-}
-
-/* Returns the 1-based line number of the start of the syntax that represents
- the token N after the current one in LEXER. Returns 0 for a T_STOP token or
- if the token is drawn from a source that does not have line numbers. */
-int
-lex_get_first_line_number (const struct lexer *lexer, int n)
-{
- const struct lex_source *src = lex_source__ (lexer);
- return src ? lex_source_next__ (src, n)->first_line : 0;
-}
-
-/* Returns the 1-based line number of the end of the syntax that represents the
- token N after the current one in LEXER, plus 1. Returns 0 for a T_STOP
- token or if the token is drawn from a source that does not have line
- numbers.
-
- Most of the time, a single token is wholly within a single line of syntax,
- but there are two exceptions: a T_STRING token can be made up of multiple
- segments on adjacent lines connected with "+" punctuators, and a T_NEG_NUM
- token can consist of a "-" on one line followed by the number on the next.
- */
-int
-lex_get_last_line_number (const struct lexer *lexer, int n)
-{
- const struct lex_source *src = lex_source__ (lexer);
- return src ? lex_token_get_last_line_number (src,
- lex_source_next__ (src, n)) : 0;
-}
-
-/* Returns the 1-based column number of the start of the syntax that represents
- the token N after the current one in LEXER. Returns 0 for a T_STOP
- token.
-
- Column numbers are measured according to the width of characters as shown in
- a typical fixed-width font, in which CJK characters have width 2 and
- combining characters have width 0. */
-int
-lex_get_first_column (const struct lexer *lexer, int n)
-{
- const struct lex_source *src = lex_source__ (lexer);
- return src ? lex_token_get_first_column (src, lex_source_next__ (src, n)) : 0;
-}
-
-/* Returns the 1-based column number of the end of the syntax that represents
- the token N after the current one in LEXER, plus 1. Returns 0 for a T_STOP
- token.
-
- Column numbers are measured according to the width of characters as shown in
- a typical fixed-width font, in which CJK characters have width 2 and
- combining characters have width 0. */
-int
-lex_get_last_column (const struct lexer *lexer, int n)
-{
- const struct lex_source *src = lex_source__ (lexer);
- return src ? lex_token_get_last_column (src, lex_source_next__ (src, n)) : 0;
+ lex_source_ofs__ (src, ofs0),
+ lex_source_ofs__ (src, ofs1));
}
/* Returns the name of the syntax file from which the current command is drawn.
must eventually free the location (with msg_location_destroy()). */
struct msg_location *
lex_get_location (const struct lexer *lexer, int n0, int n1)
-{
- struct msg_location *loc = lex_get_lines (lexer, n0, n1);
- loc->first_column = lex_get_first_column (lexer, n0);
- loc->last_column = lex_get_last_column (lexer, n1);
- return loc;
-}
-
-/* Returns a newly allocated msg_location for the syntax that represents tokens
- with 0-based offsets N0...N1, inclusive, from the current token. The
- location only covers the tokens' lines, not the columns. The caller must
- eventually free the location (with msg_location_destroy()). */
-struct msg_location *
-lex_get_lines (const struct lexer *lexer, int n0, int n1)
{
struct msg_location *loc = xmalloc (sizeof *loc);
*loc = (struct msg_location) {
- .file_name = xstrdup_if_nonnull (lex_get_file_name (lexer)),
- .first_line = lex_get_first_line_number (lexer, n0),
- .last_line = lex_get_last_line_number (lexer, n1),
+ .file_name = intern_new_if_nonnull (lex_get_file_name (lexer)),
+ .start = lex_ofs_start_point (lexer, n0 + lex_ofs (lexer)),
+ .end = lex_ofs_end_point (lexer, n1 + lex_ofs (lexer)),
+ .src = lex_source__ (lexer),
};
+ lex_source_ref (loc->src);
return loc;
}
struct lex_source *src = lex_source__ (lexer);
if (src != NULL && src->reader->error == LEX_ERROR_TERMINAL)
{
- src->head = src->tail = 0;
- src->journal_pos = src->seg_pos = src->line_pos = 0;
- src->n_newlines = 0;
+ src->length = 0;
+ src->journal_pos = src->seg_pos = 0;
+ src->n_lines = 0;
src->suppress_next_newline = false;
src->segmenter = segmenter_init (segmenter_get_mode (&src->segmenter),
false);
lex_stage_clear (&src->pp);
lex_stage_clear (&src->merge);
- lex_stage_clear (&src->lookahead);
+ lex_source_clear_parse (src);
lex_source_push_endcmd__ (src);
}
}
{
lex_stage_clear (&src->pp);
lex_stage_clear (&src->merge);
- lex_stage_clear (&src->lookahead);
+ lex_source_clear_parse (src);
for (; src != NULL && src->reader->error != LEX_ERROR_TERMINAL;
src = lex_source__ (lexer))
- lex_source_destroy (src);
+ {
+ ll_remove (&src->ll);
+ lex_source_unref (src);
+ }
}
}
\f
-static size_t
-lex_source_max_tail__ (const struct lex_source *src_)
-{
- struct lex_source *src = CONST_CAST (struct lex_source *, src_);
-
- assert (src->seg_pos >= src->line_pos);
- size_t max_tail = MIN (src->journal_pos, src->line_pos);
-
- /* Use the oldest token also. */
- struct lex_stage *stages[] = { &src->lookahead, &src->merge, &src->pp };
- for (size_t i = 0; i < sizeof stages / sizeof *stages; i++)
- if (!lex_stage_is_empty (stages[i]))
- {
- struct lex_token *first = lex_stage_first (stages[i]);
- assert (first->token_pos >= first->line_pos);
- return MIN (max_tail, first->line_pos);
- }
-
- return max_tail;
-}
-
static void
lex_source_expand__ (struct lex_source *src)
{
- if (src->head - src->tail >= src->allocated)
- {
- size_t max_tail = lex_source_max_tail__ (src);
- if (max_tail > src->tail)
- {
- /* Advance the tail, freeing up room at the head. */
- memmove (src->buffer, src->buffer + (max_tail - src->tail),
- src->head - max_tail);
- src->tail = max_tail;
- }
- else
- {
- /* Buffer is completely full. Expand it. */
- src->buffer = x2realloc (src->buffer, &src->allocated);
- }
- }
- else
- {
- /* There's space available at the head of the buffer. Nothing to do. */
- }
+ if (src->length >= src->allocated)
+ src->buffer = x2realloc (src->buffer, &src->allocated);
}
static void
{
lex_source_expand__ (src);
- size_t head_ofs = src->head - src->tail;
- size_t space = src->allocated - head_ofs;
+ size_t space = src->allocated - src->length;
enum prompt_style prompt = segmenter_get_prompt (&src->segmenter);
- size_t n = src->reader->class->read (src->reader, &src->buffer[head_ofs],
+ size_t n = src->reader->class->read (src->reader,
+ &src->buffer[src->length],
space, prompt);
assert (n <= space);
{
/* End of input. */
src->reader->eof = true;
- lex_source_expand__ (src);
return;
}
- src->head += n;
+ src->length += n;
}
- while (!memchr (&src->buffer[src->seg_pos - src->tail], '\n',
- src->head - src->seg_pos));
+ while (!memchr (&src->buffer[src->seg_pos], '\n',
+ src->length - src->seg_pos));
}
static struct lex_source *
: ll_data (ll_head (&lexer->sources), struct lex_source, ll));
}
-/* Returns the text of the syntax in SRC for tokens N0 ahead of the current
- one, through N1 ahead of the current one, inclusive. (For example, if N0
- and N1 are both zero, this requests the syntax for the current token.) The
- caller must eventually free the returned string (with free()). The syntax
- is encoded in UTF-8 and in the original form supplied to the lexer so that,
- for example, it may include comments, spaces, and new-lines if it spans
- multiple tokens. Macro expansion, however, has already been performed. */
+/* Returns the text of the syntax in SRC for tokens with offsets OFS0 through
+ OFS1 in the current command, inclusive. (For example, if OFS0 and OFS1 are
+ both zero, this requests the syntax for the first token in the current
+ command.) The caller must eventually free the returned string (with
+ free()). The syntax is encoded in UTF-8 and in the original form supplied
+ to the lexer so that, for example, it may include comments, spaces, and
+ new-lines if it spans multiple tokens. Macro expansion, however, has
+ already been performed. */
static char *
-lex_source_get_syntax__ (const struct lex_source *src, int n0, int n1)
+lex_source_syntax__ (const struct lex_source *src, int ofs0, int ofs1)
{
struct string s = DS_EMPTY_INITIALIZER;
- for (size_t i = n0; i <= n1; )
+ for (size_t i = ofs0; i <= ofs1; )
{
/* Find [I,J) as the longest sequence of tokens not produced by macro
expansion, or otherwise the longest sequence expanded from a single
macro call. */
- const struct lex_token *first = lex_source_next__ (src, i);
+ const struct lex_token *first = lex_source_ofs__ (src, i);
size_t j;
- for (j = i + 1; j <= n1; j++)
+ for (j = i + 1; j <= ofs1; j++)
{
- const struct lex_token *cur = lex_source_next__ (src, j);
+ const struct lex_token *cur = lex_source_ofs__ (src, j);
if ((first->macro_rep != NULL) != (cur->macro_rep != NULL)
|| first->macro_rep != cur->macro_rep)
break;
}
- const struct lex_token *last = lex_source_next__ (src, j - 1);
+ const struct lex_token *last = lex_source_ofs__ (src, j - 1);
/* Now add the syntax for this sequence of tokens to SRC. */
if (!ds_is_empty (&s))
{
size_t start = first->token_pos;
size_t end = last->token_pos + last->token_len;
- ds_put_substring (&s, ss_buffer (&src->buffer[start - src->tail],
- end - start));
+ ds_put_substring (&s, ss_buffer (&src->buffer[start], end - start));
}
else
{
}
static bool
-lex_source_contains_macro_call (struct lex_source *src, int n0, int n1)
+lex_source_contains_macro_call (struct lex_source *src, int ofs0, int ofs1)
{
- for (size_t i = n0; i <= n1; i++)
- if (lex_source_next__ (src, i)->macro_rep)
+ for (int i = ofs0; i <= ofs1; i++)
+ if (lex_source_ofs__ (src, i)->macro_rep)
return true;
return false;
}
The caller must not modify or free the returned string. */
static struct substring
-lex_source_get_macro_call (struct lex_source *src, int n0, int n1)
+lex_source_get_macro_call (struct lex_source *src, int ofs0, int ofs1)
{
- if (!lex_source_contains_macro_call (src, n0, n1))
+ if (!lex_source_contains_macro_call (src, ofs0, ofs1))
return ss_empty ();
- const struct lex_token *token0 = lex_source_next__ (src, n0);
- const struct lex_token *token1 = lex_source_next__ (src, MAX (n0, n1));
+ const struct lex_token *token0 = lex_source_ofs__ (src, ofs0);
+ const struct lex_token *token1 = lex_source_ofs__ (src, MAX (ofs0, ofs1));
size_t start = token0->token_pos;
size_t end = token1->token_pos + token1->token_len;
- return ss_buffer (&src->buffer[start - src->tail], end - start);
+ return ss_buffer (&src->buffer[start], end - start);
}
static void
-lex_source_error_valist (struct lex_source *src, int n0, int n1,
+lex_source_error_valist (struct lex_source *src, int ofs0, int ofs1,
const char *format, va_list args)
{
const struct lex_token *token;
ds_init_empty (&s);
- token = lex_source_next__ (src, n0);
+ token = lex_source_ofs__ (src, ofs0);
if (token->token.type == T_ENDCMD)
ds_put_cstr (&s, _("Syntax error at end of command"));
else
{
/* Get the syntax that caused the error. */
- char *raw_syntax = lex_source_get_syntax__ (src, n0, n1);
+ char *raw_syntax = lex_source_syntax__ (src, ofs0, ofs1);
char syntax[64];
str_ellipsize (ss_cstr (raw_syntax), syntax, sizeof syntax);
free (raw_syntax);
/* Get the macro call(s) that expanded to the syntax that caused the
error. */
char call[64];
- str_ellipsize (lex_source_get_macro_call (src, n0, n1),
+ str_ellipsize (lex_source_get_macro_call (src, ofs0, ofs1),
call, sizeof call);
if (syntax[0])
*m = (struct msg) {
.category = MSG_C_SYNTAX,
.severity = MSG_S_ERROR,
- .location = lex_source_get_location (src, n0, n1),
+ .location = lex_source_get_location (src, ofs0, ofs1),
.text = ds_steal_cstr (&s),
};
msg_emit (m);
lex_get_error (struct lex_source *src, const struct lex_token *token)
{
char syntax[64];
- str_ellipsize (ss_buffer (&src->buffer[token->token_pos - src->tail],
- token->token_len),
+ str_ellipsize (ss_buffer (&src->buffer[token->token_pos], token->token_len),
syntax, sizeof syntax);
struct string s = DS_EMPTY_INITIALIZER;
token->token = (struct token) { .type = T_STOP };
token->macro_rep = NULL;
token->ref_cnt = NULL;
- token->line_pos = src->line_pos;
token->token_pos = src->seg_pos;
- if (src->reader->line_number > 0)
- token->first_line = src->reader->line_number + src->n_newlines;
- else
- token->first_line = 0;
/* Extract a segment. */
const char *segment;
int seg_len;
for (;;)
{
- segment = &src->buffer[src->seg_pos - src->tail];
+ segment = &src->buffer[src->seg_pos];
seg_len = segmenter_push (&src->segmenter, segment,
- src->head - src->seg_pos,
+ src->length - src->seg_pos,
src->reader->eof, &seg_type);
if (seg_len >= 0)
break;
src->seg_pos += seg_len;
if (seg_type == SEG_NEWLINE)
{
- src->line_pos = src->seg_pos;
- src->n_newlines++;
+ if (src->n_lines >= src->allocated_lines)
+ src->lines = x2nrealloc (src->lines, &src->allocated_lines,
+ sizeof *src->lines);
+ src->lines[src->n_lines++] = src->seg_pos;
}
/* Get a token from the segment. */
for (int i = 0; i < n_lines; i++)
{
/* Beginning of line. */
- const char *line = &src->buffer[src->journal_pos - src->tail];
+ const char *line = &src->buffer[src->journal_pos];
/* Calculate line length, including \n or \r\n end-of-line if present.
- We use src->head even though that may be beyond what we've actually
- converted to tokens (which is only through line_pos). That's because,
- if we're emitting the line due to SEG_END_COMMAND, we want to take the
- whole line through the newline, not just through the '.'. */
- size_t max_len = src->head - src->journal_pos;
+ We use src->length even though that may be beyond what we've actually
+ converted to tokens. That's because, if we're emitting the line due
+ to SEG_END_COMMAND, we want to take the whole line through the
+ newline, not just through the '.'. */
+ size_t max_len = src->length - src->journal_pos;
const char *newline = memchr (line, '\n', max_len);
size_t line_len = newline ? newline - line + 1 : max_len;
}
const struct lex_token *t = lex_stage_nth (&src->pp, ofs);
- size_t start = t->token_pos;
- size_t end = t->token_pos + t->token_len;
const struct macro_token mt = {
.token = t->token,
- .syntax = ss_buffer (&src->buffer[start - src->tail], end - start),
+ .syntax = ss_buffer (&src->buffer[t->token_pos], t->token_len),
};
const struct msg_location loc = lex_token_location (src, t, t);
n_call = macro_call_add (mc, &mt, &loc);
.token = expansion.mts[i].token,
.token_pos = c0->token_pos,
.token_len = (c1->token_pos + c1->token_len) - c0->token_pos,
- .line_pos = c0->line_pos,
- .first_line = c0->first_line,
.macro_rep = macro_rep,
.ofs = ofs[i],
.len = len[i],
Returns true if successful, false on failure. In the latter case, SRC is
exhausted and 'src->eof' is now true. */
static bool
-lex_source_get_lookahead (struct lex_source *src)
+lex_source_get_parse (struct lex_source *src)
{
struct merger m = MERGER_INIT;
struct token out;
&out);
if (!retval)
{
- lex_stage_shift (&src->lookahead, &src->merge, 1);
+ lex_source_push_parse (src, lex_stage_take_first (&src->merge));
return true;
}
else if (retval > 0)
.token = out,
.token_pos = first->token_pos,
.token_len = (last->token_pos - first->token_pos) + last->token_len,
- .line_pos = first->line_pos,
- .first_line = first->first_line,
/* This works well if all the tokens were not expanded from macros,
or if they came from the same macro expansion. It just gives up
};
if (t->ref_cnt)
++*t->ref_cnt;
- lex_stage_push_last (&src->lookahead, t);
+ lex_source_push_parse (src, t);
for (int i = 0; i < retval; i++)
lex_stage_pop_first (&src->merge);
static void
lex_source_push_endcmd__ (struct lex_source *src)
{
- assert (lex_stage_is_empty (&src->lookahead));
+ assert (src->n_parse == 0);
+
struct lex_token *token = xmalloc (sizeof *token);
*token = (struct lex_token) { .token = { .type = T_ENDCMD } };
- lex_stage_push_last (&src->lookahead, token);
+ lex_source_push_parse (src, token);
+}
+
+static void
+lex_source_push_parse (struct lex_source *src, struct lex_token *token)
+{
+ if (src->n_parse >= src->allocated_parse)
+ src->parse = x2nrealloc (src->parse, &src->allocated_parse,
+ sizeof *src->parse);
+ src->parse[src->n_parse++] = token;
+}
+
+static void
+lex_source_clear_parse (struct lex_source *src)
+{
+ for (size_t i = 0; i < src->n_parse; i++)
+ lex_token_destroy (src->parse[i]);
+ src->n_parse = src->parse_ofs = 0;
}
static struct lex_source *
lex_source_create (struct lexer *lexer, struct lex_reader *reader)
{
+ size_t allocated_lines = 4;
+ size_t *lines = xmalloc (allocated_lines * sizeof *lines);
+ *lines = 0;
+
struct lex_source *src = xmalloc (sizeof *src);
*src = (struct lex_source) {
+ .n_refs = 1,
.reader = reader,
.segmenter = segmenter_init (reader->syntax, false),
.lexer = lexer,
+ .lines = lines,
+ .n_lines = 1,
+ .allocated_lines = allocated_lines,
};
lex_source_push_endcmd__ (src);
return src;
}
-static void
-lex_source_destroy (struct lex_source *src)
+void
+lex_set_message_handler (struct lexer *lexer,
+ void (*output_msg) (const struct msg *,
+ struct lexer *))
+{
+ struct msg_handler msg_handler = {
+ .output_msg = (void (*)(const struct msg *, void *)) output_msg,
+ .aux = lexer,
+ .lex_source_ref = lex_source_ref,
+ .lex_source_unref = lex_source_unref,
+ .lex_source_get_line = lex_source_get_line,
+ };
+ msg_set_handler (&msg_handler);
+}
+
+void
+lex_source_ref (const struct lex_source *src_)
+{
+ struct lex_source *src = CONST_CAST (struct lex_source *, src_);
+ if (src)
+ {
+ assert (src->n_refs > 0);
+ src->n_refs++;
+ }
+}
+
+void
+lex_source_unref (struct lex_source *src)
{
+ if (!src)
+ return;
+
+ assert (src->n_refs > 0);
+ if (--src->n_refs > 0)
+ return;
+
char *file_name = src->reader->file_name;
char *encoding = src->reader->encoding;
if (src->reader->class->destroy != NULL)
free (file_name);
free (encoding);
free (src->buffer);
+ free (src->lines);
lex_stage_uninit (&src->pp);
lex_stage_uninit (&src->merge);
- lex_stage_uninit (&src->lookahead);
- ll_remove (&src->ll);
+ lex_source_clear_parse (src);
+ free (src->parse);
free (src);
}
\f
lex_string_read,
lex_string_close
};
+\f
+struct substring
+lex_source_get_line (const struct lex_source *src, int line)
+{
+ if (line < 1 || line > src->n_lines)
+ return ss_empty ();
+
+ size_t ofs = src->lines[line - 1];
+ size_t end = line >= src->n_lines ? src->length : src->lines[line];
+ return ss_buffer (&src->buffer[ofs], end - ofs);
+}
projects / pspp / blobdiff